Guide Tools Fake Wan VACE Video Nudify Workflow Tutorial

[Deleted member 21996425]

How much Vram you got?

hm, idk. The difference in my workflow shouldn't matter too much. Id also check the version of the nodes ur using -- I ran into issues with those sometimes. I have like 12gb of vram

 

[Mra742]

Change the model within the model loader to be a general i2v wan 2.1 model. Right now it's using a VACE model which isn't supposed to be used with the I2V workflow. Also feel free to remove the Low Noise Self Forcing Lora Loader, I forgot to remove that in the workflow.

You can find the GGUF i2v wan 2.1 models here.

city96/Wan2.1-I2V-14B-480P-gguf · Hugging Face

We’re on a journey to advance and democratize artificial intelligence through open source and open science.

huggingface.co

city96/Wan2.1-I2V-14B-720P-gguf at main

We’re on a journey to advance and democratize artificial intelligence through open source and open science.

huggingface.co

Bro, I changed the model. It was a big change, but now the result is blurry. Do you know what the problem is? Is there a specific lora to remove the clothes? I'm using the one recommended in the other workflows.

 

[dvh]

Is there a way to run wan 2.2 with 12gb vram or can I only run version 2.1?

 

[thenyeguyfi]

Does anyone know how to smooth out the transitions and improve the consistency between segments of video?

Sometimes the body/skin gets darker or the lighting changes dramatically between segments. It varies by video, even with the same settings + prompt and I don't know what causes this.

Also there is sometimes an issue of clothes/bra/bra straps progressively reappearing after like 9+ seconds. They look transparent like an x-ray or molten into the skin.

Another thing, sometimes when the person turns around, the masked preview shows a mask over the clothing as if it recognizes it correctly, but the generated video does not reflect that, and the rear clothing is not removed. Then when they face toward the camera again, the mask preview is again masking their clothing properly, but nothing is generated over it.

I'm using singingminer's wan 2.2 v4 workflow on runpod with default loras and default settings. The only issue I see with it is that my standard GGUF loader is not set up properly so I have to bypass it and use multigpu loader.

Based on my experience, (not an expert), but the skin getting darker was due to using the 2.2 general nsfw lora, when I went back to the 2.1 version (using 2.2 workflow) the skin didnt darken as much.

The clothes re-appearing for me has been because there's one frame where the clothes starts appearing, usually if you section that out for masking, it covers it for me

 

[jacksonbaby]

The chest looks too fake, is there any way to improve it

 

[Jerker901210]

lsmtatk_wan22_inpaint_v4.1_test2.json
I tried this new workflow, but encountered some problems after running it. What is the reason for this

TO ANYONE HAVING CLIP LOADER GGUF ERROR:
Older version will work.

> Click 'Manager' in top-right
> Custom nodes in workflow
> Find gguf and click switch ver.
> Select 2.4.4, install
> Restart and refresh

Fixes standard clip loader node being broken.
You can also just use the Multi-GPU clip loader by right-clicking and bypassing the standard one.